Effective intervention programs must be based on salient variables that improve diabetes control. Although literature supports that family relational processes influence diabetes control in adolescence and suggests that autonomy development is a major process associated with self-care, little attention has focused on autonomy as a primary capability that ungirds self- care and diabetes control as adolescence progresses. The long- term objective of this research is the facilitation of autonomy development in adolescents with IDDM to enhance self-care and diabetes control as adolescent's progress through early and middle adolescence. Specific aims are (1) to describe the developmental progression of family relational processes, autonomy, self-care and diabetes control in adolescents with IDDM and (2) to evaluate a model of influences of adolescent autonomy development and family relational processes on self-care and diabetes control, in which autonomy development has a central mediating role. A combined cross-sectional and longitudinal cohort-sequential design will be used and 274 adolescents between the ages of 11 and 15 years and their families will be enrolled and followed prospectively for 2 years until ages 13 through 17. The focus will be on the parent- adolescent subsystem and data will be obtained during a home visit through self-report questionnaires and a revealed difference family interaction task. The task will involve discussion of adolescent and diabetes specific family issues by parent(s) and adolescent. Measures of adolescent autonomy (emotional, cognitive and behavioral), family relational processes (cohesion, hostile conflict, parental separation anxiety, autonomous-relatedness, and inhibition of autonomy), self care (universal, illness-related and division of diabetes responsibility), and diabetes control will be employed. Diabetes control will be by glycosylated hemoglobin levels at two intervals four months apart. Cross sectional data will be analyzed using regression/correlational procedures. Longitudinal data will be analyzed by hierarchical linear models (HLM) with random coefficients. Model fit will be evaluated by several methods, including goodness of fit and diagnostic plots of individual subject slopes.